Great Circle Associates List-Managers
(November 1995)

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: linked vs autonomous list servers
From: Wes Morgan <morgan @ engr . uky . edu>
Date: Thu, 2 Nov 1995 10:51:56 -0500
To: list-managers @ greatcircle . com

>But the philosophy of the LISTSERV system is that ALL list servers 
>are linked together, they share information on what lists are available.
>Majordomo, on the other hand, and SmartList, do not share this linking 
>So there are these two philosophies about mailing lists, as reflected in these 
>various software implementations.

I wouldn't read *too* much into the differing approaches.  LISTSERV
grew up in the BITNET world; BITNET was a tree-structured network, in
which sites could only talk to their nearest neighbors.   (Due to the
lack of a direct link, BITNET traffic between UKentucky and ULouisville
once traversed *8* sites!)  Keep in mind, too, that BITNET did not use
routers _per se_; each site along the way had to actively process each
transaction. In that environment, it made perfect sense for the LISTSERVs 
to communicate as peers; since every site was acting as a router for 
someone else, this peered approach was far more effective than routing 
all those individual requests across the net.

Over in the uucp and nascent TCP/IP Internet world, however, the picture
was different.  It was a simple matter for a TCP/IP site (and even some
uucp sites) to connect directly to the destination site and drop the mail;
intermediate sites would not be bothered by this.  Internet sites did not
face the traffic management problem suffered by BITNET; once placed on
the network, the routers - instead of the actual intermediate sites - han-
dled traffic flow.  Given that environment, peering listservers would 
actually *create* a middleman!

I don't perceive these differences as philosophical.  I see them as the
natural reactions to the network topologies/technologies in use at the
time.  For tree-structured BITNET, peering servers was a *HUGE* win; for
the TCP/IP Internet, peering would actually add steps to the process.
Each of these decisions was a no-brainer in terms of network load and
burden on participating sites.

It will be interesting to see how the LISTSERV model works out on the
TCP/IP Internet in the long term.  Personally, I foresee the larger
lists benefitting from the peering approach; I don't know if smaller
lists will see *that* much of a windfall.


Indexed By Date Previous: Re: Enough is enough
From: srb @ cuci . nl (Stephen R. van den Berg)
Next: Re: Enough is enough
From: Mitch Collinsworth <mkc @ graphics . cornell . edu>
Indexed By Thread Previous: Re: Enough is enough
From: "Roger Fajman" <RAF @ CU . NIH . GOV>
Next: Re: linked vs autonomous list servers
From: Eric Thomas <ERIC @ SEARN . SUNET . SE>

Search Internet Search